Summary
Prognostic models of liver transplantation (LT) for hepatocellular carcinoma (HCC) mainly derive from LT cohorts with numerous hepatitis C virus (HCV) patients. The AFP model, which is currently used in France to select LT candidates, was derived from a cohort of LT performed between 1988 and 2001, including a majority of HCV‐positive recipients. The emergence of new direct‐acting antiviral therapies and subsequent decrease of HCV incidence may change the generalizability of such models. We performed an external validation of the AFP model in a cohort of recipients transplanted between 2005 and 2018. Although multivariable analysis identified all three model’s factors (AFP level, largest tumor size, number of nodules) as predictors of tumor recurrence, the AFP model showed poor discrimination and calibration in the present cohort. This poor performance could be related to significant differences between the derivation and the present cohort in terms of etiology, severity of underlying liver disease, tumor burden and differentiation, and use of neoadjuvant treatments. The present findings suggest that the decline of HCV‐induced HCC among LT candidates may compromise the generalizability of the AFP model in more recent LT cohorts. Further studies are required for updating or building more robust prognostic models.